How to continue train, If it is stopped somehow?

[OliverLING071]

No description provided.

[knicholes]

If you read main.py, you'll see all of the command line arguments you can supply. One of those arguments is --resume. Note, however, that if you haven't trained at least 100,000 epocs, you will not have saved your model, but there will be one saved in your results dir under YOUR_DATASET_NAME_params_latest.py. If you copy that into your results/YOUR_DATASET_NAME/model/YOUR_DATASET_NAME_params_<zero padded integer of the iteration you last saved on (a multiple of 1,000 by default), you can use the --resume, and it'll pick up where it left off.